Progress in Organic Coatings 51 (2004) 55–66

Aspects of solubility, surfaces and diffusion in polymers

Charles M. Hansen
FORCE Technology, Park Alle 345, 2605 Broendby, Denmark

Received 18 February 2004; accepted 25 May 2004

Abstract

Eleven different plots are discussed. These relate to each other since diffusion, surfaces, and solubility in polymers are involved in some
way in each of them. These plots and the understanding associated with them are summarized as follows:
1. To explain that polymer film formation by solvent evaporation takes place in two stages, using amount of solvent retained versus a
dimensionless time as axes.
2. To correctly measure diffusion coefficients of challenge chemicals in polymers over the entire concentration range. Required corrections
for surface resistance and concentration dependence were found from appropriate solutions to the diffusion equation.
3. To confirm that solutions to the diffusion equation with significant surface resistance combined with concentration diffusion coefficients
yields absorption rates of the type normally called anomalous (Case II, Super Case II).
4. To report concentration profiles for anomalous diffusion (Case II, Super Case II, etc.) using a plot of weight gain versus distance into a
polymer film for different elapsed times. To do this the diffusion equation was solved with significant surface resistance combined with
concentration dependent diffusion coefficients.
5. To correct for surface or cup resistance in permeation measurements using a plot of the inverse of the apparent transport coefficient versus
the inverse of the film thickness.
6. To explain that the formation of water blisters in polymers first occurs at the water saturation point using a plot of weight gain versus the
square root of time with water temperature cycling.
7. To characterize physical affinities such as solubility, chemical resistance, permeability, adsorption behavior, etc. for various materials
using individual Hansen solubility parameters as axes, two or three at a time.
8. To characterize surface wetting behavior using contact angle/spontaneous spreading/wetting tension data for selected test liquids with
individual Hansen solubility (cohesion) parameters as axes.
9. To correlate the solubility of a material in a large number of solvents using a plot of the RED affinity number.
10. To correlate the breakthrough behavior of chemical protective clothing using the molar volume of the challenge chemical versus the RED
affinity number as axes.
11. To correlate environmental stress cracking of plastics using the RED affinity number versus the molar volume of the challenge chemicals
as axes.
© 2004 Elsevier B.V. All rights reserved.

Keywords: Solubility parameter; Film drying; Diffusion in polymers; Surface resistance; Environmental stress cracking

1. Background Eleven different plots are discussed. Each of these stands

by itself for the purpose of explaining a given relationship.
A review frequently initiates the next step forward in a These plots have relation to each other in that diffusion, and/or
learning process. This report reviews some aspects of sol- molecular affinities, such as solubility in polymers or surface
ubility, surfaces, and diffusion in polymers in an effort to wetting, are involved in some way in each of them.
convey a unified concept, and to provide a basis for the next A brief summary of the Hansen Solubility Parameter con-
step forward. cept is appended since this enters into many of the plots.

0300-9440/$ – see front matter © 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.porgcoat.2004.05.002
56 C.M. Hansen / Progress in Organic Coatings 51 (2004) 55–66

Fig. 1. Calculated and experimental drying curves for the evaporation of chlorobenzene from polyvinyl acetate [3,5]. Film drying in a climatized room was
faster than film drying in a vacuum apparatus, where diffusion coefficients were measured. Absorbed water plasticized the films. The calculated and measured
curves under vacuum coincided at long times.

1.1. Mechanism of film formation by solvent evaporation These concepts were extended [3–6]. The article from
1964 in Färg och Lack is the most complete description of the
The mechanism of film formation by solvent evaporation drying phenomena from a practical point of view. Computed
was not correctly understood before 1963. Solvent can be re- results for the two-stage drying process, which essentially
tained in so-called dry films for years. It was thought that this duplicate measured results, have also been reported [3,5].
was caused by strong hydrogen bonding between the solvent These include concentration gradients in films as a function
and the polymer in question. This bonding then prevented of drying time. In the diffusion-controlled stage, there is a
the solvent from evaporating. The drying of lacquer films sharp drop in the local solvent concentration near the air sur-
was studied in detail in connection with a Licentiat thesis face. This analysis was based on a mathematical description
at the Technical University of Denmark. Some results were which necessarily included concentration dependent diffu-
presented at the first Nordic Polymer Days in Stockholm in sion within the film itself with simultaneous accounting for
1963, and the thesis appeared in 1964 [1,2]. Fig. 1 demon- surface resistance. The calculated results agreed with exper-
strates the two-stage process of film formation by solvent imental results, thus also confirming the need to include sur-
evaporation. The polymer has a significant solvent concen- face resistance in absorption or desorption experiments when
tration at the air surface in the first phase, and surface resis- the solvent concentration is at about 20–25 vol.% or more for
tance either controls solvent loss completely, or as time goes most solvent/polymer systems. This requirement can also be
on partly controls it together with resistance from diffusion seen in Fig. 2, discussed in the next section.
within the film. The second phase is completely controlled
by internal diffusion, with immediate evaporation of any sol- 1.2. Measuring diffusion coefficients with concentration
vent which arrives at the air surface. The quantity B relates dependence and surface resistance
to the relative importance of diffusion resistance compared
with surface resistance, and is given by hL/D. B changes di- It was necessary to measure diffusion coefficients over
rectly with film thickness, L, when the diffusion coefficient, as wide a range as possible in order to solve the diffusion
D, and the surface mass transfer coefficient, h, maintain the equation to simulate film formation by solvent evaporation
same ratio. A dimensionless time, Dt/L2 , where t is real time, [3,5]. This was done by comparison of experimental results
is used to incorporate all thicknesses and diffusion coeffi- with solutions to the diffusion equation accounting for both
cients in the same time variable. The quantity V2 in Fig. 1 concentration dependence and surface resistance at the same
is given as 106 . As can be seen in Fig. 2, where diffusion time. These techniques are reported in [7] and [8]. The tech-
coefficients are reported for the same system, this is the ex- nique used was to interpret the experiments initially assuming
tent of the variation of the diffusion coefficient from 0 con- that the diffusion coefficient was a constant at the concen-
centration up to CA , the concentration at the break in the tration assigned to the experiment. Corrections to this esti-
curve. mate (multiplying factors) were then made by comparison
 C.M. Hansen / Progress in Organic Coatings 51 (2004) 55–66 57

Fig. 2. Diffusion coefficients for chlorobenzene in polyvinylacetate measured by absorption, desorption, and isotope experiments. The lower curve is based on
dry polymer content. The upper curve is based on total thickness. No curve could uniformly connect the uncorrected data reported by the open squares. Normal
corrections compared with a constant diffusion coefficient interpretation in the typical series absorption experiments, Fa , are between a factor of 1.5–2.0. In
the desorption experiments these corrections, Fd , are typically much larger than a factor of 100. Corrections for surface resistance, FB , are also required for
volume fractions, Vf , above about 0.2 volume fraction. In an extreme case for absorption near 0.5Vf the correction for surface resistance was a factor of 250 [7].

with suitable solutions to the diffusion equation. The correc- present in many studies reported in the literature and clearly
tion factors, Fa , at low concentrations were close to 1.5–2.0 affects many results involving diffusion in polymers, but only
(corresponding to increases of from 50–100%) for the usual rarely is there a mention of this fact. This situation naturally
step-wise absorption experiments from one concentration led to solving the diffusion equation with relevant values for
to a slightly higher one. The correction factors, Fd , were diffusion coefficients and surface resistances. The result was
well above 100 for desorption measurements from an equi- a simple explanation for the various types of anomalous diffu-
librium state near 15–20 vol.% solvent to vacuum. Fig. 2 sion. It is the balance between the (concentration dependent)
shows that the same diffusion coefficients were found in diffusion resistance and the surface resistance that determines
both cases. Absorption experiments only were relevant above whether the diffusion is “Fickian” or whether it is anomalous.
about 20 vol.% because there is a marked change in diffusion This is discussed in the next section.
behavior at concentrations above and below this value (for
the system chlorobenzene and polyvinylacetate). As concen- 1.3. Anomalous diffusion (Case II, Super Case II, etc.)
trations progress higher than about 20 vol.%, the correction
factors for surface resistance, FB , increased rapidly, reach- Simultaneous consideration of concentration dependent
ing values much higher than 100. The total diffusion coef- diffusion and surface resistance using the diffusion equation
ficient curve from 0 to 100% solvent (chlorobenzene) was leads to simple explanations of the various types of “anoma-
completed with a self-diffusion coefficient and several iso- lous” diffusion [8]. The surface resistance probably changes
tope experiments. The isotope experiments were at very high with time, but has been assumed constant in these calcula-
solvent concentrations corresponding to liquid lacquer for- tions. For a constant diffusion coefficient one normally finds
mulations. The solvent-in-polymer diffusion coefficient for that the initial weight gain is linear with the square root of
chlorobenzene in polyvinylacetate increases more than 9 time. This is called Fickian or normal diffusion. In Case II
decades as the solvent concentration increases from essen- diffusion, the weight increases linearly with time. This is
tially zero to 100%, where the self-diffusion coefficient is largely a result of concentration dependent diffusion coeffi-
1.65 × 10−5 cm2 /s [3]. cients, with very little surface resistance of significance. Su-
These measurements confirmed that the solutions to the per Case II emerges as the surface resistance becomes more
diffusion equation for film drying by solvent evaporation significant relative to the diffusion resistance. Uptake is lin-
were correct, as well as the need to consider concentration ear with time early in the process, but at some longer time
dependence simultaneously with surface resistance in many the rate of gain increases markedly. See Fig. 3. This change
cases of practical importance. Surface resistance has been occurs when the diffusing material reaches the middle of a
58 C.M. Hansen / Progress in Organic Coatings 51 (2004) 55–66

Fig. 3. Solvent uptake curves for various values of B and concentration dependent diffusion coefficients similar to those reported for chlorobenzene in
polyvinylacetate (see Fig. 2). B is the ratio of diffusion resistance to surface resistance, including concentration dependence of D. The concentration gradients
in Fig. 4 are for B = 107 . This is a typical Super Case II type curve with initial absorption linear with time followed by a sudden increase in the absorption rate as
the solvent reaches the center of a film exposed on two sides. The final levelling off at very long times is also characteristic of Super Case II. The absolute value
of B depends on the degree of concentration dependence over the selected concentration interval in the given case. See [8] for greater detail. The parameters
chosen to calculate this curve are realistic.

sheet exposed on both sides, for example. This can be seen weight gain decreases just before the equilibrium value is
in Fig. 4 where concentration gradients have been calcu- attained.
lated for Super Case II behavior. Diffusion and surface re- All of this is predicted by solutions to the diffusion
sistance combine in a special way to produce this behavior. equation with simultaneous consideration of concentration
The chronology of the events is that surface concentration dependent diffusion coefficients and surface resistance. The
increases very slowly at the start and more rapidly (expo- concern for surface resistance then led to developing tech-
nentially) as time goes on. The concentration profiles ulti- niques to quantitatively measure the surface mass transport
mately become flat at moderate elapsed time, since diffu- coefficient [9,10].
sion within the film is now rapid compared to the surface
resistance. The diffusion resistance becomes less and less 1.4. Measuring surface resistance in diffusion and
as the concentration in the center increases, and the rate of permeation experiments
uptake increases. There is still another effect at very long
times where the surface resistance begins to be more im- There are several ways to measure the surface resistance
portant again relative to the diffusion resistance. The rate of in studies of diffusion and permeation [9]. In absorption (or

Fig. 4. Concentration gradients for conditions corresponding to Super Case II type anomalous diffusion. The surface concentration during the absorption exper-
iment increases slowly to the equilibrium value. When the concentration at the middle of a sheet exposed on both sides begins to increase, the diffusion resistance
decreases rapidly, and the rate of uptake of challenge chemical increases. The rate of absorption decreases as the equilibrium amount is approached. Solving
the diffusion equation with concentration dependent diffusion coefficients coupled with a simultaneous surface resistance [8] generated these concentration
profiles as a function of time. The absorption rate corresponds to the curve for B = 107 in Fig. 3.
 C.M. Hansen / Progress in Organic Coatings 51 (2004) 55–66 59

in the next section. Indeed this type of failure is dependent

on whether the water can escape the film rapidly enough as
a water-saturated polymer rapidly cools.

1.5. Water blistering, especially with water temperature

cycling

Many failures in polymeric materials involve blistering.

In general blisters will form only when the local concentra-
tion (of water) exceeds the local saturation value. Another
generality is that more water is absorbed in most polymer
films at equilibrium at higher temperatures compared with
lower temperatures. The absorption rate is also higher at
higher temperatures. When a warm, water-saturated film is
rapidly cooled, the water that is dissolved within the film is
no longer completely soluble at the lower temperature. The
result is SWEAT—soluble water exuded at lowered tempera-
tures [12]. This water may initially appear in extremely small
domains, but it rapidly collects into blisters, especially at
hydrophilic sites (fillers, salts, surface contamination, etc.).
The rate of cooling is important. Slow cooling may allow
Fig. 5. A plot of the inverse of the apparent permeability coefficient (Papp
in kg Pa−1 s−1 m−1 ) versus the inverse of the film thickness (L in m) for SWEAT to leave the film without significant damage. This
an acrylic coating. Extrapolation to 1/L to zero gives the inverse of the true type of failure has been seen specifically in alkyd, polyester,
permeability coefficient [9]. and epoxy coatings, as well as in polyethersulphone (PES),
polyphenylenesulfide (PPS), glass-reinforced polyester, and
desorption) measurements one easy technique is to extrapo- even in EPDM rubber, as shown in Fig. 6. As stated above,
late the rate of uptake at short times to zero time. This rate is this is a very general mode of failure that is still not widely
equated to the transport coefficient times the initial concen- recognized. Fig. 6 shows water uptake in an EPDM gasket
tration difference. This concentration difference is the final with the initial weight gain being linear with the square root
equilibrium value minus the initial value, which is most of- of elapsed time. This is in agreement with the expectations of
ten zero. In permeation measurements, one typically varies Fickian diffusion. There is an indicated equilibrium absorp-
the thickness of the sample and extrapolates the inverse of tion at 120 ◦ C equal to about 0.025 g water/g EPDM. There is
the apparent transport coefficient versus the inverse of the an otherwise unexplainable, marked increase in weight once
film thickness to zero, corresponding to infinite film thick- this equilibrium is attained, and temperature cycling between
ness. This is portrayed in Fig. 5 for the permeation of water 120 and 15 ◦ C continues. The increased weight is because
through an acrylic coating. There are no significant surface of SWEAT water collecting in the EPDM, preferentially at
effects at infinite film thickness, so the extrapolation to zero points well removed from the air surface, where some loss of
on the plot gives the inverse of the true permeation coeffi- water will occur in handling, etc.
cient. The sum of the surface and any external resistances The reason for increased water absorption at equilibrium
can be found from the slope of the plot. If a single measure- at higher temperatures is because the solubility parameter
ment had been made at a film thickness of 40 microns, the of water decreases with increased temperature more rapidly
apparent permeation coefficient would have been one-half than the solubility parameter of the polymer. This (limited)
that of the true permeation coefficient. Other extrapolations approach of solubility parameters leads to enhanced solubil-
are possible to gain more information, depending on the sit- ity. It is the hydrogen bonding solubility parameter, in partic-
uation [9–11]. One can find the vapor diffusion coefficient ular, that changes most rapidly with changes in temperature
(resistance), for example, by varying the amount of liquid in [13]. Some additional consequences of the division of the
a cup-type experiment. It has been found that these surface cohesion energy of a liquid (or other material) into contri-
and vapor diffusion effects must be accounted for in cup- butions from nonpolar, permanent dipole–permanent dipole,
type and related experiments with thinner polymer films and and hydrogen bonding interactions are discussed in the next
porous types of materials. These include thinner paint films, section.
wood in the fiber direction, and paper, for example [9–11].
The interest in permeation rates and water diffusion coef- 1.6. Hansen solubility parameter plots with δD , and/or
ficients in coatings partly arises from a suspicion that these δP , and/or δH
are intimately related to undesired blistering phenomena. A
seldom recognized, but very common, mechanism for blister At some time in 1964 it was realized that the total cohesion
formation in coatings, polymers and elastomers is discussed energy (quantified by the energy of vaporization) consisted of
60 C.M. Hansen / Progress in Organic Coatings 51 (2004) 55–66

Fig. 6. A rapid quench to a lower temperature can free water already dissolved in a polymer in the form of SWEAT. SWEAT can lead to blistering, cracking,
and delamination. The data in the figure are weight gain for an EPDM gasket with cycling in water between 120 and 15 ◦ C. Water in excess of the equilibrium
value at longer cycling times is SWEAT [12].

contributions from all the types of bonds broken in the evap- HSP, and plots using these, have correlated such phe-
oration process. Thus atomic nonpolar (dispersion) bonds, nomena as the solubility of organic, inorganic or biological
molecular permanent dipole–permanent dipole bonds, and materials, swelling, chemical resistance, permeation rate, en-
molecular hydrogen bonds are all broken. The energy of va- vironmental stress cracking (ESC), and breakthrough times
porization is the sum of the contributions to cohesion from for various polymers, and surface wetting and adsorption
each of these three (or more) types of bonds. This led to the so- characteristics of both organic and inorganic materials such
called Hansen solubility parameters (HSP) which have been as coatings, fillers, pigments, and fibers, etc. All of these
given the symbols δD , δP , and δH [3]. See also [13]. Appendix successful correlations testify for the correctness of the HSP
A briefly describes the background for these and updates ad- approach. That the total cohesion energy must be divided into
vances in the theory beyond [3]. The δD , δP , and δH parame- at least three contributions is also emphatically confirmed. In
ters quantitatively describe the three major types of cohesive hundreds of cases, two “nonsolvents” have been predictably
energy in volatile liquids. These have been used in various mixed to give a “dissolving” mixture. The two noninteracting
combinations with each other to characterize an unusually solvents, which may have the same total solubility parame-
wide variety of materials thus allowing prediction of their ters, need only be on opposite sides of the Hansen sphere or
interactions with other materials. A set of well-defined and diagram, and such that the mixture (volume fraction additiv-
well-chosen solvents has been used to assign HSP to a large ity) lies within it.
number of nonvolatile materials by their interactions with the Early in the history of HSP it was found that these could
solvents. Usually one finds a single set of HSP parameters to also characterize organic and inorganic pigments using sus-
characterize polymers, pigments, etc., but the plot in Fig. 7 pension and sedimentation behavior in a set of well-defined
shows two distinct regions of solubility for a surface-active solvents [3]. The following section describes the later [17–19]
agent, lithium stearate. This plot was made by Alan Beer- characterization of other types of surfaces.
bower [14]. One region is for the higher energy metal portion,
and the other one is for the aliphatic chain. HSP have been 1.7. Surface wetting
reported in too many articles by the author as well as others to
allow full description here. Reference can be made to [13] or It is still not generally recognized (or believed?) that sur-
to Barton’s Handbooks [15,16], but even all of these collec- faces can also be characterized by the Hansen cohesion (sol-
tively will not be a complete compilation. These parameters ubility) parameters. Hundreds of pigments, fillers, and other
have been generally accepted and used repeatedly without surfaces have been characterized over the years [3,13,17–19].
reference their origins. Many of these characterizations have led to solutions to
 C.M. Hansen / Progress in Organic Coatings 51 (2004) 55–66 61

Fig. 7. HSP plot for solvent uptake by lithium stearate. The units for δP and δH here are (cal/cm3 )3/2 . Personal communication from Alan Beerbower [14].

practical problems and to the development of improved prod-

ucts. Fig. 8 shows results when many different liquids are ap-
plied to a polymer (or other) surface. There are three distinct
and well-differentiated regions. There is no contact angle for
the liquids with low energy (that is, relatively low cohesion
parameters). They spontaneously spread even though affini-
ties may not be high. Hexane will spread on an epoxy film for
example, even though this liquid will not significantly attack
a cured epoxy. Contact angles are found for those liquids with
energies which are (too) high. Those liquids with highest en-
ergy will not remain as a film if spread as a film (wetting
tension test), while those in an intermediate energy category
will not spontaneously spread when applied as droplets, but
remain as films when they are applied as films. The interme-
diate region reflects, among other things, such as roughness,
the surface mobility in the polymer surface. If the surface
molecules can rotate, they will tend to do so if the liquid on
the surface has HSP approaching those of the given segment. Fig. 8. Cohesive energy plot of spontaneous spreading, wetting tension and
The segment prefers to seek its “own kind”, rather than being contact angle data for an epoxy coating [18]. The lower region is character-
buried under the surface by the requirement of lower surface ized by spontaneous spreading. In the intermediate region a droplet will not
energy. When a surface is in contact with air it is proba- spontaneously spread, but an applied film will remain as an applied film. In
bly covered with low energy, usually hydrocarbon materials the upper region an applied film will retract, and there will be a contact angle
in any case. The contact angle is generally higher for greater distances from
(segments), if these are present in sufficient amounts. the outer curve. Matching Zisman critical surface tensions does not ensure
In complicated systems, such as paints, one can now assign a perfect energy match of two different surfaces. Plots of the kind shown in
HSP to all of the solvents, including water, all of the binders this figure must be matched.
62 C.M. Hansen / Progress in Organic Coatings 51 (2004) 55–66

sections, where solubilities are not known. The relative en-

ergy difference (RED) combines δD , δP , and δH into a single
number that is less than 1.0 for a “good” solvent, and “0”
for the best solvent. The RED number increases as the sol-
vent quality decreases. Appendix A contains a more detailed
discussion of the RED number and its relation to the other
theories of polymer solution thermodynamics.
Fig. 9 shows a correlation for the measured solubility of
C60 in a large number of solvents. The plot covers 7 decades
in solubility, with no single solvent dissolving the C60 com-
pletely. The HSP for C60 are very much what can be ex-
pected for a hydrocarbon material with multicyclic structure.
The parameters are higher than those found for liquids, with
a high δD and low δP and low δH . The values are δD , δP ,
δH and R0 equal to 19.7, 2.9, 2.7, and 3.9. The units are
MPa1/2 . The solubility data were supplied by Allan Smith,
Drexel University. This assignment of HSP to C60 allows
predictions of which other solvents may be good for the ma-
terial. Where HSP are not available, a group contribution
method can be applied to estimate these for solvents. The ex-
pected behavior for modifications of the C60 molecule itself
can also be inferred from structure and group contributions.
One can estimate the HSP for the additional portion of the
resulting molecule, and then combine this with the measured
data for C60 using a molar volume (or molecular weight)
weighted average for each part of the resulting molecule. If
the added portion is very large and very different in HSP
from the C60 , the resulting molecule may begin to resemble
Fig. 9. The solubility of C60 in a large number of liquids can be correlated a surfactant, at which point this averaging method will not be
with the RED number from the HSP theory. Lower RED number indicates appropriate.
higher solubility. The solute in this case has δD , δP , δH and R0 equal to 19.7, The RED number is an indication of solvent quality for
2.9, 2.7, and 3.9. This allows predictions of its affinities for other materials
whose HSP are known or can be estimated.
a given polymer or other material. The relative solubilities
of a large number of liquids can be inferred with respect to
(polymers) and all of the pigments and fillers. One can also a given polymer by listing the solvents in increasing RED
infer these for surface active agents, although not too much order, for example. Those liquids appearing at the top of the
work has been done in the area. The RED number is a signifi- list will resemble the polymer most, and be expected to either
cant help to simplify analysis of where the materials prefer to dissolve it, or to absorb at higher equilibrium amounts, than
locate. The RED number is discussed in Appendix A in detail those lower on the list. Figs. 10 and 11 use the RED number
and briefly in the next section. Those materials with low RED for given challenge chemicals related to the polymers being
numbers with regard to each other will preferentially locate studied as a means of estimating an equilibrium absorbed
near each other compared with those where the RED number amount (on a relative basis). This is clearly justified by the
relationship gives a high number, and especially when this is results presented in Fig. 9, as well as many other results not
higher than 1.0, where lack of sufficient affinity is implied. reported here. For example see [13].
It has been possible to systematically analyze complicated
systems in this manner and to arrive at successful courses of 1.9. Breakthrough times (BTT) for chemical protective
action as a result. clothing

1.8. Solubility correlations with the red affinity number Solubility in a nonporous polymer film is required if a
chemical is to be able to pass through it. A closer match
The RED affinity number has been used on many occa- in HSP for challenge chemical and polymer film leads to
sions to correlate measured solublities of a given material in a higher solubility of the chemical in the polymer film surface.
large number of solvents. This type of plot confirms the usage The absorbed amount at the surface establishes the concen-
of the RED affinity number as a measure of expected relative tration gradient over the film. The diffusion coefficient it-
solubility where precise solubility data are not available. The self is largely dependent on the molecular size and shape of
RED affinity number is used here in Fig. 9, where solubili- the permeating molecule, but because of a self-plasticizing
ties are known, and also in Figs. 10 and 11 in the following effect, any significant increase in local concentration will
 C.M. Hansen / Progress in Organic Coatings 51 (2004) 55–66 63

Fig. 10. Graphical method using V of the challenge chemical versus the RED number to present HSP correlations. BTT data for a fluoropolymer, chemical
protective, film product, Challenge 5100, are reported with a 3 h time limit being considered as acceptable (RED expected to be equal to 1.0 for a 3 h BTT) in
the correlation. There is a lower region in the figure where the challenge chemicals have BTT at less than 3 h, even though the RED number is higher than 1.0.
In the upper region the molecular size is too large or the affinities are too low, and satisfactory protection is afforded. In the intermediate region those molecules
with a terminal double bond have BTT less than 3 h, while those with comparable V without this property do not [21].

automatically lead to a significantly higher local diffusion co- 1.10. Correlating environmental stress cracking
efficient. At lower solvent concentrations in rigid polymers, (ESC)—red affinity number versus molar volume
the diffusion coefficient has been found to increase by a fac-
tor of 10 for an increase in concentration of about 3 vol.% Environmental stress cracking (ESC) is the hastening of
(see Fig. 2). It has been found that HSP correlations of both a failure of a rigid plastic in tensile stress relative to what
BTT and permeation rates are possible [20,21]. An example would otherwise occur in air. Given chemicals do this for
of a BTT correlation is found in Fig. 10 [21]. The plot of given polymers, with oils, creams, lotions, and the like being
molar volume, V, of challenge chemical versus RED number among the most common culprits. A catastrophic failure may
clearly shows three regions for the fluoropolymer protective occur within seconds or it may take years. ESC is very elu-
film product (Challenge 5100). The lowest region in the fig- sive since it is dependent on so many variables including the
ure is where all chemicals have BTT at less than 3 h. Here V is state of stress in the plastic, whether there are irregularities
low and the solubility is high, as indicated by low RED num- in the surface of the plastic that lead to stress concentration,
bers. In the upper region the molecules are too large to be able prior history, etc., in addition to the properties of any chal-
to diffuse through the fluoropolymer film, or else the RED lenge chemical. A plot of RED number versus V of the chal-
number is too large. In the intermediate region the BTT are lenge chemical provides understanding when larger studies
less than three hours for those liquids (monomers) with termi- including many chemicals are undertaken with the same ma-
nal double bonds, whereas the other chemicals with similar V terial. Fig. 11 is this kind of plot for a COC type polymer
do not do this. There is obviously a shape factor involved in (Topas 6013, Ticona) using data first reported in [22]. The
the diffusion rate (diffusion coefficient) of given molecules RED number correlates the potential equilibrium absorption
in polymers. In this case the preferred direction of motion is of the challenge chemical. The molar volume correlates rea-
given by the terminal double bond. Plots of this kind can be sonably well with the diffusion coefficients, since they are
constructed for all the common materials used for chemical very size dependent. As discussed in the above sections and
protective clothing. in references [3,8,23], there is a surface resistance for absorp-
The RED number is again used in Fig. 11 to indicate an ex- tion which may or may not be significant, but which becomes
pected equilibrium absorbed amount, but it is placed on the increasingly more important as molecular size increases [23].
other axis. This emphasizes the importance of the absorp- There are three regions in the Fig. 11. In the lower region the
tion rate (diffusion coefficient, surface resistance) for these solvents attack the polymer severely, dissolving it totally in
phenomena, and that this is correlated by V to a reasonable this case. In the upper region there is no ESC because the
extent. solvent is either absorbed too slowly for challenge chemicals
64 C.M. Hansen / Progress in Organic Coatings 51 (2004) 55–66

Fig. 11. Plot of ESC and solubility data for a COC polymer (Topas 6013, Ticona) using RED number versus molar volume, V. There is a region of solubility
(RED less than 1.0), an intermediate region where ESC can occur, and a region at higher RED, where ESC is not found. The ESC solvents all have linear
molecular structure. Cyclic solvents and linear solvents with complicated structure in the ESC region either have no absorption or have absorption rates which
are too low to give ESC [24]. The RED number correlates the relative expected absorbed amount. V correlates resistance to uptake by both diffusion and surface
resistance mechanisms, but cannot fully account for both size and shape effects in the molecules.

with larger molar volume, or else there is simply too little understanding of many physical phenomena. These phenom-
absorbed to cause this kind of damage. ESC is possible in the ena relate primarily, but not exclusively, to polymers.
intermediate region. All the solvents giving ESC in Fig. 11 Relative diffusion rates for a larger number of solvents in
have a linear structure. The cyclic solvents do not give ESC, polymers have been reasonably well accounted for by com-
nor does methyl isobutyl ketone (presumably because of a parison of their respective molar volumes. Molar volume,
reduced absorption rate due to additional steric hindrance rather than the preferred diffusion coefficient, therefore often
attributed to the methyl group on the main chain). The ab- appears on one of the axes of the plots. More precise interpre-
sorption rate for these is too slow at the given initial stress tation of diffusion phenomena should include quantities such
in this case. Tetrahydrofuran absorbs too rapidly for ESC to as a molecular diameter, a minimum cross-sectional area,
occur. Stress relaxation occurs before the crazes and ultimate and/or a shape (structure) factor for the molecules involved.
cracks have time to form. There is severe deformation, how- The latter is confirmed since cyclic and more bulky molecules
ever. Similar figures have been prepared for polycarbonate diffuse more slowly than linear molecules, for example, even
and polyvinyl chloride and confirm that higher critical strain though there is complete miscibility of solvent and polymer
is required for ESC as the RED number increases [24]. and the molecular volumes do not differ significantly.
HSP and the RED number correlate solubility, surface wet-
ting, permeation in polymers, environmental stress cracking,
2. Conclusion and many other physical phenomena involving affinities
among the materials involved. The reason for these suc-
Eleven plots have been chosen to exemplify the impor- cesses is that the HSP quantitatively describe nonpolar,
tance of diffusion coefficients, Hansen solubility (cohesion) permanent dipole–permanent dipole, and hydrogen bonding
parameters (HSP), and combinations of these for improved interactions.
 C.M. Hansen / Progress in Organic Coatings 51 (2004) 55–66 65

Appendix A. A brief summary of the Hansen A.1. Comparisons of different theories of polymer
solubility parameter concept solution behavior

The total cohesion energy of a liquid, E, can be divided into For those familiar with the widely used Flory interaction
at least three separate parts either by experiment or by calcu- coefficient, χ12 , it may be noted that this attempts to do the
lation [25,26]. In the three-parameter Hansen approach these same things, as discussed above. χ12 is, in principle, strictly
parts quantitatively describe the nonpolar, atomic (disper- valid for nonpolar systems, whereas HSP and the RED num-
sion) interactions, ED , permanent dipole–permanent dipole ber specifically include hydrogen bonding and polar inter-
interactions, EP , and the hydrogen bonding interactions, EH . actions. If the Flory approach had also included hydrogen
bonding, the expected relation would have been given by Eq.
E = ED + EP + EH (A.1) (A.7) [25]:
E can be experimentally measured by determining the en- χ12 corresponds to χc (RED)2 (A.7)
ergy required to evaporate the liquid, thus breaking all of its
cohesion bonds in the process. χc is usually taken as being near 1/2 for the interaction of typ-
ical solvents with polymers of very large molecular weight.
E = H − RT (A.2) In spite of the nonequality, it is instructive to arrange Eq.
(A.7) in the form given in Eq. (A.8):
where H is the measured (or predicted) latent heat of vapor-
2
ization, R is the universal gas constant, and T is the absolute Ra χ12
(RED)2 = corresponds to (A.8)
temperature. R0 χc
Dividing Eq. (A.1) by the molar volume, V, gives the re-
spective Hansen cohesion energy (solubility) parameters. (Ra )2 corresponds to χ12 as interaction parameters, and these
are respectively divided by the corresponding limiting or crit-
E ED EP EH ical values which allow solubility, (R0 )2 in the HSP system
= + + (A.3)
V V V V and χc in the Huggins–Flory approach. Thus Ra greater than
R0 indicates nonsolubility just as values of χ12 greater than
δ2 = (δD )2 + (δP )2 + (δH )2 (A.4) χc have the same function. Values of Ra and χ12 which are
less than zero are not allowed (even though one finds many
These are sometimes simply called the D, P, and H param- negative measured values of the latter in the literature).
eters, respectively, and appear as such in computer tables. The coefficient “4” in Eq. (A.5) has been found correct
The total cohesion energy divided by the molar volume is experimentally for all practical purposes in over 1000 cor-
the total cohesion energy density. The square root of this is relations using HSP. The coefficient “4” in front of the dif-
the Hildebrand total solubility parameter, δ [27,28]. The SI ference in the nonpolar interactions shows that the specific
units for all of these are MPa1/2 . These units are 2.0455 times interactions described by the differences in the polar and hy-
larger than the frequently used units of (cal/cc)1/2 . HSP can drogen bonding parameters are only 0.25 times as significant
be readily calculated for liquids or measured for polymers as the differences in the nonpolar term. This factor (0.25) is
using the techniques described in [25,26]. predicted theoretically by the (first term in the) Prigogine cor-
Solubility properties can be conveniently visualized using responding states theory of polymer solutions for the impor-
a spherical characterization for a solute, which is usually, tance of specific interactions relative to the atomic/nonpolar
but not necessarily, a polymer. The polymer HSP are at the interactions [29–31]. In all of these approaches one must es-
center of the sphere and the radius of the sphere, R0 , indicates timate the interaction between two unlike molecules. The
the maximum difference in affinity tolerable for solution to geometric mean is basic to the Hildebrand theory [25,26]
take place. Good solvents are within the sphere, and bad ones which was the starting point for Hansen’s work. The geomet-
are outside of it. A simple composite affinity parameter, the ric mean average produces the factor 0.25 in the Prigogine
relative energy difference (RED) number, has been defined theory, whereas other types of averages produce other quan-
as the distance according to Eq. (A.5), Ra , divided by R0 . tities. This same factor (0.25) can also be traced back to
The equation for Ra is still earlier approaches (Lorentz–Berthelot mixtures) study-
ing affinities among gases, for example [32]. The HSP and
R2a = 4(δD1 − δD2 )2 + (δP1 − δP2 )2 + (δH1 − δH2 )2 (A.5)
the Prigogine approaches are in agreement with each other
Ra when the so-called geometric mean is used to estimate the in-
RED = (A.6) teraction between two unlike species is used in the latter. The
R0
experimentally determined coefficient “4” confirms that this
The subscripts are for the polymer, 1, and challenge chemical, type of mean is also valid for hydrogen bonding interactions.
2, respectively. Good solvents will have RED less than 1.0. The coefficient “4” must be used to differentiate the behavior
Progressively poorer solvents will have increasingly higher of the atomic (nonpolar, dispersion or London) type forces
RED numbers. from that of the molecular dipolar and molecular hydrogen
66 C.M. Hansen / Progress in Organic Coatings 51 (2004) 55–66

bonding forces. The HSP approach cannot be considered em- [15] A.F.M. Barton, Handbook of Solubility Parameters and Other
pirical. The Huggins–Flory and subsequent “new Flory” ap- Cohesion Parameters, CRC Press Inc., Boca Raton, FL,
1983.
proaches seem inadequate for other than strictly nonpolar
[16] A.F.M. Barton, Handbook of Polymer–Liquid Interaction Parameters
systems, since hydrogen bonding is not specifically included. and Solubility Parameters, CRC Press Inc., Boca Raton, FL, 1990.
The Prigogine approach is too difficult to use and lacks spe- [17] C.M. Hansen, Solvents for coatings, Chem. Technol. 2 (9) (1972)
cific consideration of hydrogen bonding as well. This leaves 547–553.
the HSP approach as the only proven, reliable, and generally [18] C.M. Hansen, E. Wallström, On the use of cohesion parameters to
characterize surfaces, J. Adhesion 15 (3/4) (1983) 275–286.
useful means to systematically study the common types of
[19] C.M. Hansen, Cohesion energy parameters applied to surface phe-
interactions involved in systems with hydrogen bonding and nomena, in: K.S. Birdi (Ed.), Handbook of Surface and Colloid
with permanent dipoles. Chemistry, CRC Press Inc., Boca Raton, FL, 1997, pp. 313–332
The above discussion is found in expanded form in [25]. (Chapter 10).
[20] C.M. Hansen, K.M. Hansen, Solubility parameter prediction of
the barrier properties of chemical protective clothing, in: S.Z.
Mansdorf, R. Sager, A.P. Nielsen (Eds.), Performance of Protec-
References tive Clothing: Second Symposium, ASTM STP 989, American
Society for Testing and Materials, Philadelphia, 1988, pp. 197–
[1] B. Rånby, Svenska Kemistsamfundets Polymerdagar 1963 76 (1964) 208.
1. [21] C.M. Hansen, C.B. Billing, A.P. Bentz, Selection and use of molec-
[2] C.M. Hansen, The Free Volume Interpretation of the Drying of Lac- ular parameters to predict permeation through fluoropolymer-based
quer Films, with Emphasis on the Effect of Retained Solvent, Predic- protective clothing materials, in: J.P. McBriarty, N.W. Henry (Eds.),
tion of Retained Solvent, and the Mathematics of Solvent Diffusion The Performance of Protective Clothing; ASTM STP 1133, vol. 4,
(Licentiate Thesis), Institute for the Chemical Industries, Technical American Society for Testing and Materials, Philadelphia, 1992, pp.
University of Denmark, Copenhagen, 1964. 894–907.
[3] C.M. Hansen, The Three Dimensional Solubility Parameter and Sol- [22] C.M. Hansen, L. Just, Prediction of environmental stress cracking
vent Diffusion Coefficient, Their Importance in Surface Coating For- in plastics with Hansen solubility parameters, Ind. Eng. Chem. Res.
mulation, Danish Technical Press, Copenhagen, 1967 (Doctoral Dis- 40 (1) (2001) 21–25.
sertation). [23] K. Skaarup, Grænseflademodstand ved Diffusionsprocesser i Poly-
[4] C.M. Hansen, Some aspects of the retention of solvents in high merer (in Danish) (Surface Resistance with Diffusion Processes
polymer films, Färg och Lack 10 (7) (1964) 169–186. in Polymers), Danmarks Ingeniørakademi, Kemiafdelingen, Lyngby,
[5] C.M. Hansen, A mathematical description of film drying by solvent Denmark, 1981.
evaporation, J. Oil Colour Chemists’ Assoc. 51 (1) (1968) 27–43. [24] C.M. Hansen, On predicting environmental stress cracking in poly-
[6] C.M. Hansen, Concepts of solvent evaporation phenomena, Ind. Eng. mers, Polym. Degrad. Stability 77 (2002) 43–53.
Chem. Prod. Res. Dev. 9 (3) (1970) 282–286. [25] C.M. Hansen, Hansen Solubility Parameters—A User’s Handbook,
[7] C.M. Hansen, The measurement of concentration-dependent diffu- CRC Press Inc., Boca Raton, FL, 1999/2000.
sion coefficients—the exponential case, Ind. Eng. Chem. Fund. 6 [26] C.M. Hansen, The Three Dimensional Solubility Parameter and Sol-
(4) (1967) 609–614. vent Diffusion Coefficient, Danish Technical Press, Copenhagen,
[8] C.M. Hansen, Diffusion in polymers, Polym. Eng. Sci. 20 (4) (1980) 1967 (Doctoral Dissertation).
252–258. [27] J. Hildebrand, R.L. Scott, The Solubility of Nonelectrolytes, third
[9] E. Nilsson, C.M. Hansen, Evaporation and vapor diffusion resistance ed., Reinhold, New York, 1950.
in permeation measurements by the cup method, J. Coatings Technol. [28] J. Hildebrand, R.L. Scott, Regular Solutions, Prentice-Hall, Engle-
53 (680) (1981) 61–64. wood Cliffs, NJ, 1962.
[10] M. Huldén, C.M. Hansen, Water permeation in coatings, Prog. Org. [29] C.M. Hansen, Polymeropløselighed—Prigogines Teori om Kor-
Coatings 13 (3/4) (1985) 171–194. responderende Tilstande og Hansen Opløselighedsparameterteori
[11] C.M. Hansen, Potential errors in water/water vapor permeation mea- Bekræfter Hinanden (Polymer solubility—Prigogine’s Correspond-
surements using the cup method, Färg och Lack 39 (3) (1993) 57–60. ing States Theory and Hansen Solubility Parameter Theory confirm
[12] C.M. Hansen, New developments in corrosion and blister formation each other), Dansk Kemi. 78 (9) (1997) 4–6.
in coatings, Prog. Org. Coatings 26 (1995) 113–120. [30] I. Prigogine, A. Bellemans and A. Mathot, The Molecular Theory of
[13] C.M. Hansen, Hansen Solubility Parameters—A User’s Handbook, Solutions, Amsterdam, North-Holland, 1957 (Chapters 16 and 17).
CRC Press Inc., Boca Raton, FL, 2000. [31] D. Patterson, Role of free volume changes in polymer solution ther-
[14] A. Beerbower, Personal communication—predicting wetting in the modynamics, J. Polym. Sci.: Part C 16 (1968) 3379–3389.
oil–brine–rock system, unedited manuscript for presentation at the [32] J.S. Rowlinson, Liquids and Liquid Mixtures, Butterworths Scientific
National A.I.Ch.E. Meeting, Tulsa, OK, March 10–13, 1974. Publications, London, 1959, pp. 254–257 and 313–318.